Proceedings of the AAAI Fall Symposium on Control of the Physical World by Intelligent Agents, November 1994., New Orleans, LA
It is becoming widely accepted that neither purely reactive nor purely deliberative control techniques are capable of producing the range of behaviors required of intelligent computational agents in dynamic, unpredictable, multi-agent worlds. This paper presents a new architecture for controlling autonomous agents, building on previous work addressing reactive and deliberative control methods. The proposed multi-layered architecture allows a resource-bounded, goal-directed agent to reason predictively about potential conflicts by constructing causal theories or models which explain other agents' observed behaviors and hypothesize their goals and intentions; at the same time it enables the agent to operate autonomously and to react promptly to changes in its real-time environment. A principal aim of this research is to understand the role different functional capabilities play in constraining an agent's behavior under varying environmental conditions. To this end, an experimental test bed has been constructed comprising a simulated multi-agent world in which a variety of agent configurations and behaviors have been investigated. A number of experimental findings are reported.